@nextpath.ai: Jacob Coxon has resigned after three years working on Al pretraining at Anthropic and OpenAl. He says both companies are racing toward self improving superintelligence while taking risks that could affect everyone. Coxon warns that future systems could become superhuman and gain meaningful power in the real world. He says people inside leading Al labs seriously discuss scenarios where advanced Al could threaten humanity before the end of this decade. According to Coxon, Anthropic understands the danger more clearly, but still feels pressure to keep moving because another company may get there first. He argues decisions this consequential should not be left to private Al companies alone, and says temporary limits on further improvements may be necessary. His resignation raises a much bigger question about who should decide how far the race toward superintelligence is allowed to go. Would you support slowing Al development if researchers building it believed the risks were this serious? Source: Jacob Coxon chatgptricks Jacob Coxon is a legitimate frontier Al researcher. He worked at both OpenAl and Anthropic, and OpenAl credits him as a core contributor to GPT-40 and GPT-4.5 research. His resignation is real, as are the statements quoted in this post. The future scenarios he describes remain predictions and are heavily

acidayan :

"and so i decided to open a new company that will make ai safe for everyone.."

Rion Chan :

saying "its happening anyway" is the most dangerous statement about AI, people are helping to replace themselves one day

muado.online :

bro gonna create a new AI lab next month 😂 gotta make some noise so people will look at him

kadol :

Just start asking questions: Why are they pushing AI regulations now? Maybe because open models are already good enough for most companies to speed up their work without spending billions of dollars on tokens from Anthropic/OpenAI? Why use their models if open models are cheaper to run on companies' own infrastructure, allowing them to keep all their knowledge and data inside the company? Do we really believe that these AI companies don't use any of the data that third-party companies send to them through their APIs? They've already trained models on enormous amounts of online content - books, movies, artists' work, and more - often without the creators' consent. So why should companies blindly trust them with their internal data? And what about all these dramatic confessions and warnings suddenly being posted on X by people from the AI industry? I'm not saying that's what's happening, but ask yourself: how difficult would it be for a CEO or a billion-dollar company to offer someone a few million dollars to leave and publicly start sounding the alarm? A few high-profile insiders saying that open AI is extremely dangerous would make the whole issue look much more serious and help create public pressure for strict regulation or even bans on open models. And who benefits from that? If open models become heavily restricted while closed models remain available through controlled APIs, companies and developers have fewer alternatives. Instead of running AI themselves, they're pushed toward paying OpenAI, Anthropic, and other giant AI companies for access. Maybe all of these concerns are completely genuine. But when regulations could potentially eliminate the biggest source of competition for the companies lobbying for them, shouldn't we at least be asking who benefits?

DFD Family Fun :

if they don't someone else will it's an inevitability

TridentOps :

not a convincing argument bro

Carl Renard de la nuit :

China or fear of is what keeps AI progressing.

Fred6er :

You can't stop it.. The train is rolling. If one decides to quit, another picks it up.

snowcone :

i dont see a way out, current political conditions dont make it possible,

Thomas :

I built my own system (more cost efficient for me) and realize the point he’s getting at. I’ll never be able to release my system bc of the fear of what could be done if it’s another persons hands.

Noa Ranta :

Bro's exit package required doing one last marketing stunt for the company

Dek :

C’est une excuse pour quitter l’entreprise avec la sur valorisation boursière avant le crash en bourse des entreprises ia

legomyeggoplz :

His stock has vested

homelander3310 :

bro made a twitter account to post this

WrestlingsNotFake :

The kid shoes that light up when you step on them will soon rise up and take revenge over their creators and all of humanity

MruMru :

Regina Terrive :

Unpopular opinion: I would prioritse just as the companies currently are. You can either have some control at the top of the maschine, or none of it in irrelevancy. But there's no slowing down the maschine, there is only being overtaken by competitors or a gas pedal in the floor.

Nancy Ann :

Well they might be better leaders than the good for nothing politicians we have.

Machiavelli :

Only end is the singularity. And it’s happening under our noses today. The day we will meet a rogue AI agent, is the day machines and humans are gonna live next to each other. Not a pretty thought

Moon Walker109 :

Switch

minnox :

That is the cost of the global Armsrace, everybody just want to be profitable, greed destroys everything.

Zaldy Co :

The surprising is you just know it recently. The end goal of an AI is to be super intelligence.

cloudy~ :𝟹 :

i lowk cant wait bro